Lifespan varies widely among individuals, yet the extent to which such variation persists when genetic and environmental differences are minimized remains unclear. Here we quantify such stochastic lifespan variation in a naturally clonal vertebrate and test whether and how this variation is linked to early-life behavioral individuality. We followed N = 33 genetically identical Amazon mollies ( Poecilia formosa ), separated on day 1 of their life into highly standardized environments, from birth to death. Despite genetic uniformity and environmental standardization, lifespan varies markedly, spanning 502 – 826 days. Continuous high-resolution behavioral tracking during the first four weeks of life reveals that seemingly stochastic early-life activity differences explain 32.5% of this variation. Higher activity predicts shorter lifespan during the first two weeks, but as activity levels and among-individual variation in activity decline over early development, a U-shaped relationship emerges, with both low- and high-activity individuals outliving those with intermediate activity. These findings show that signatures of lifespan emerge within days of birth, even among genetically identical individuals, highlighting developmental stochasticity and early-life contingencies as major contributors to variation in life-history outcomes.

Optimal stocking density allows shrimp farmers to maximise profit while maintaining animal welfare. To date, holistic assessments of chronic crowding stress effect on Litopenaeus vannamei are lacking regarding biological impairments, stress-related behaviors, and recovery capacity. Twelve tanks of a recirculation aquaculture systems were maintained at three stocking densities: 1 kg/m2 (low density-Low), 2 kg/m2, (standard density-Standard) and 4 kg/m2 (high density-High). The experiment was divided into phase 1 (21 days stress) and phase 2 (21 days recovery). After stress, shrimp at Low treatment exhibited the highest survival rate, best growth performance, antennae and uropods quality. Stress behavior included more frequent abnormal swimming patterns and loss of balance. Hemolymph analysis suggested homeostatic mechanisms for the mobilization of energy from storage organs. Mild upregulation of specific markers indicated cellular/oxidative stress. Limited immune system effects indicate low correlation with crowding or distress. During the recovery phase, mortalities ceased and partial compensatory growth was recorded. Shrimp recovered from injuries, while abnormal behaviors frequency also decreased. Oxidative stress marker expression returned to baseline levels for unstressed shrimp. Morphological and behavioral observations can be used industrially to assess crowding stress or as tools for developing welfare indices or monitoring systems.

Understanding and modeling animal behavior is essential for studying collective motion, decision-making, and bio-inspired robotics. Yet, evaluating the accuracy of behavioral models still often relies on offline comparisons to static trajectory statistics. Here we introduce a reinforcement-learning-based framework that uses a biomimetic robotic fish (RoboFish) to evaluate computational models of live fish behavior through closed-loop interaction. We trained policies in simulation using four distinct fish models-a simple constant-follow baseline, two rule-based models, and a biologically grounded convolutional neural network model-and transferred these policies to the real RoboFish setup, where they interacted with live fish. Policies were trained to guide a simulated fish to goal locations, enabling us to quantify how the response of real fish differs from the simulated fish's response. We evaluate the fish models by quantifying the sim-to-real gaps, defined as the Wasserstein distance between simulated and real distributions of behavioral metrics such as goal-reaching performance, inter-individual distances, wall interactions, and alignment. The neural network-based fish model exhibited the smallest gap across goal-reaching performance and most other metrics, indicating higher behavioral fidelity than conventional rule-based models under this benchmark. More importantly, this separation shows that the proposed evaluation can quantitatively distinguish candidate models under matched closed-loop conditions. Our work demonstrates how learning-based robotic experiments can uncover deficiencies in behavioral models and provides a general framework for evaluating animal behavior models through embodied interaction.

ABSTRACT The study of among-individual phenotypic variation arising in the apparent absence of genetic and environmental differences has recently emerged as a rapidly growing research area. Despite growing recognition of its existence and fitness relevance, it remains unknown whether signatures of such presumably stochastically induced variation can be transmitted across generations. To address this knowledge gap, we performed a two-generation behavioral screening with a naturally clonal fish: 34 genetically identical mothers and their 232 offspring were separated after birth into near-identical environments, with early-life behavior being tracked continuously at high resolution, constituting a total of ∼19,000 observation hours. We find consistent among-individual differences in behavior (i.e., activity and feeding patterns) in both mothers and offspring. Mother behavior correlates with offspring activity (but not offspring feeding): mothers that spend more time feeding produce more active offspring. We find no evidence of body size (maternal or offspring) mediating mother-offspring behavioral associations. Our study provides first evidence for the non-genetic transmission of among-individual phenotypic differences that arise in the apparent absence of genetic or environmental differences, highlighting the potential importance of this variation for evolutionary processes and the adaptive potential of populations.

Predator-prey interactions can drive arms races of complex adaptation and response patterns. While collective behaviour is common as anti-predator behaviour and well understood mechanistically, little is known on predator strategies in the context of collective prey defences. Here, we investigated attack strategies of predatory birds confronted with an effective collective anti-predator behaviour employed by shoals of small freshwater fish (Poecilia sulphuraria). Analysing over 700 attacks revealed that predators face a trade-off: placing attacks in the shoal centre leads to a stronger prey response, which decreases attack frequency but increases chance for success. The causal relationship between attack location and prey-response plasticity was confirmed experimentally through simulated attacks. Predators adjusted their strategies with conspicuous species avoiding the centre, sacrificing optimal attack positions for lower prey responses. A cryptic predator causing overall low-intensity responses favoured central attacks, showing that the magnitude of the prey's defence is driving predator attack strategies. Additionally, we report priming effects where repeated attacks in spatial proximity intensified the prey's response, leading to predator species spacing attacks farther apart, suggesting priming counters sequential hunting strategies. We highlight the complex behavioural patterns underlying predator-prey dynamics in the wild, while mechanistic aspects of collective priming prompt interesting future research directions.

Biomimetic robots are innovative tools for guiding animal behavior by triggering avoidance responses that steer movement away from the robot. Potential applications include livestock management and environmental conservation, such as directing fish away from pollution and oil spills. While predator-like replicas have traditionally been used to provoke avoidance, less is known about how conspecific-like robotic platforms may elicit threat-like responses depending on their approach behavior. To address this gap, we programmed a conspecific-like robot to repeatedly approach live fish in a free-swimming setup, with varying approach speeds in each trial. Our results show that repeated exposure increased the likelihood of avoidance responses, indicating changes in behavioral reaction over time. In some instances, fish displayed threat-associated responses, including freezing and evasive maneuvers characterized by extreme turns and accelerations relative to baseline swimming. The initiation of these evasive events depended on both robot–fish distance and relative speed, suggesting that avoidance responses in this context are not determined by distance alone. Additionally, avoidance speed increased dynamically with both robot speed and proximity. These findings provide quantitative insights into how approach dynamics shape avoidance behavior toward a conspecific-like robot, informing future models of fish interactions and the design of robotic systems aimed at guiding fish movement through controlled aversive cues.

Paternal effects, i.e., effects of fathers on the phenotype of their offspring that are not mediated by the transmission of alleles, are increasingly recognized as a potentially significant source of phenotypic variation across taxa - even in the absence of paternal care. Gynogenetic systems, which rely on sperm to trigger embryogenesis without incorporating male genetic material, provide a powerful way to experimentally isolate paternal effects from effects caused by the integration of male genetic material. Up to now, however, paternal effects remain largely unexplored in these systems. Here, we tested for non-genetic paternal effects in the gynogenetic Amazon molly (Poecilia formosa): a naturally clonal, all-female species with no parental care. Using a highly controlled breeding experiment involving 60 Atlantic molly males (Poecilia mexicana) and 54 Amazon molly females from a single clonal lineage, we generated 128 broods and 2,435 offspring. While males were drawn from a naturally variable stock population, females - next to being genetically identical - were standardized for age, size, descent, and developmental experience. We asked whether male identity or body size predicted offspring size - a key offspring phenotypic trait. We also asked whether male identity or body size predicted brood size. Male identity explained either no or only very small proportions of the variation in offspring or brood size. Larger males were weakly associated with larger offspring, but this effect was minimal (partial R2 ~ 1.5%). However, these patterns did not hold consistently across all data exclusion criteria and analytical variants, underscoring their tentative nature and highlighting the need for further investigation. Our study offers one of the first empirical tests of male effects in a gynogenetic vertebrate, providing valuable quantitative benchmarks for the magnitude of such effects in gynogenetic systems.

Animal collectives can exhibit striking synchrony in behavior even when members differ consistently among each other when alone. While synchrony in collective movements or signaling as found in schooling fish or flashing fireflies is well-studied, these behaviors typically do not require individuals to compromise physiological needs when adjusting direction, speed, or timing, as they are generally almost cost-free to produce. In contrast, less is known about how individuals coordinate state-dependent behaviors shaped by internal physiological needs that may limit their ability to conform. We address this question by combining agent-based modeling with empirical observations of a distinctive case of synchronization in fish - the collective air-breathing of juvenile Arapaima gigas. We show that individuals differ consistently in surfacing rhythms when alone, yet in a large shoal of about 200 same-aged individuals, a substantial portion of the group surfaces within the same second. Our analysis, supported by individual-based simulations of inherently non-periodic coupled oscillators (units that act stochastically in isolation), reveals a simple social interaction rule by which synchrony emerges despite individual variation in surfacing rhythms. The model, matched to our empirical data, suggests that assortative social responsiveness ("cluster synchrony") can buffer internal constraints, enabling coordination without overriding individual physiological limitations.

In order to better understand Litopenaeus vannamei aquaculture practices within the burgeoning European shrimp aquaculture sector, a 17-point questionnaire was developed and distributed to industrial or scientific shrimp experts based in or affiliated with Europe. Although shrimp aquaculture is not yet widespread in Europe, 29 responses were received. Recirculating aquaculture systems (RAS) and biofloc are the predominant production systems, often featuring alarm notifications. Daily system checks and animal health assessments are common, with records for behaviour and physical health monitoring, regardless of system type. Enrichment strategies, like artificial substrates and feeding trays, are frequently used. Exuviae management is debated and it is removed manually or automatically, although shrimps also consume them. Optimal water conditions identified include temperatures of 24-30 degrees C, pH between 7.0 and 8.5, salinity of 15-25 ppt, and alkalinity of 51-200 mg/L CaCO3. Nitrogen species, alkalinity, and CO2 levels are recommended to be checked a few times a week. Neither optimal water parameters nor their check frequency correlate with the system type used. Health issues reported include antennal, uropodal, rostral, exoskeletal, muscular, and eye impairments. Gill quality seems better in RAS, while antennae and eye status may relate to stocking density. Common abnormal behaviours include swimming abnormalities, escape behaviour, lack of feeding, lethargy, and muscle cramps. Experts recommend developing health indices, ethograms, artificial intelligence monitoring, and management guidelines to advance the sector. In lieu of existing official standards, these findings offer partial guidelines and recommendations for indoor shrimp aquaculture, focusing on water chemistry, shrimp welfare, and management practices. Moulting dynamics, exuvia management, handling techniques, enrichments, and mineral requirements remain key questions for the sector.

Studies on collective cognition provide many examples of how the efficient spread of information within groups leads to benefits with increasing group size. However, little is known if groups also amplify maladaptive information such as false alarms and whether such costs reduce possible benefits. Here, we investigated wild fish shoals responding collectively with escape dives when attacked by birds. We analyzed the collective response in reaction to bird attacks and similar but harmless flybys as a function of shoal size. Larger shoals increasingly detected predator attacks (i.e., true positives), while their response facing harmless flybys (i.e., false alarms) remained constant. Furthermore, decision time decreased with increasing shoal size. Larger shoals were thus able to simultaneously overcome two major trade-offs inherent in solitary decision-making: the trade-off between true and false positives and the trade-off between speed and accuracy. Our findings set the stage for the next generation of studies investigating the mechanisms underlying collective decision-making.

Developmental plasticity at the behavioral repertoire level allows animals to incrementally adjust their behavioral phenotypes to match their environments through ontogeny. Quantifying this plasticity in sufficient resolution across substantial periods of development, however, has been challenging. Here, we use high-resolution tracking to monitor 45 genetically identical Amazon mollies (Poecilia formosa) reared in near-identical environments over their first four weeks of life. We analyze behavior at 0.2-s resolution to assess plasticity across entire behavioral repertoires. Testing a key prediction from Bayesian models-that plasticity should decline in stable environments-we measure plasticity using both individual behavioral metrics and a bespoke "behavioral entropy" approach in a multi-dimensional phenotype space. Surprisingly, and despite closely conforming to model assumptions, we find a consistent initial two-week increase in movement plasticity before a decline. These results challenge expectations about how plasticity unfolds early in life and highlight the importance of continuous behavioral tracking for evaluating developmental theories.

Animal collectives are capable of performing behaviors with high degrees of synchrony though their members might differ consistently and substantially in the focal behavior when alone. It is thus not entirely understood how these consistent differences in behavior at the individual level can be (socially) integrated into synchronized behaviors at the collective level. Here we show an unprecedented synchronized behavior in fish - the collective air-breathing of juvenile Arapaima gigas . Individuals of this obligate air-breathing fish from South America differed in their time between consecutive breaths when recorded alone in an aquaculture facility. However, when together in a shoal of about 200 same aged individuals, breathing is executed by a substantial portion of the shoal - within the same second. Our analysis of the individual and collective breathing patterns supported by stochastic individual-based simulations of inherently non-periodic coupled oscillators revealed that this degree of collective synchronization could be achieved by having some kind of assortative interaction rules where individuals respond towards one cluster/subgroup members stronger than to other cluster/subgroup members. By integrating this cluster synchrony rule we could successfully simulate highly synchronized collective behavior with varying proportions of otherwise diverse individuals taking part, matching our experimental observations and providing a mechanism to synchronize agents that differ consistently in the behavior in focus when in isolation. ### Competing Interest Statement The authors have declared no competing interest. Deutsche Forschungsgemeinschaft, https://ror.org/018mejw64, BI1828/2-1, BI1828/3-1, Excellence Strategy – EXC 2002/1 “Science of Intelligence” – project number 390523135

Paternal effects, i.e., effects of males on the phenotypes of their offspring that are not caused by the integration of male genetic material, are increasingly recognized as a potentially significant source of phenotypic variation across taxa - even in the absence of paternal care. Gynogenetic systems, which rely on sperm to trigger embryogenesis without incorporating male genetic material, provide a powerful way to experimentally isolate potential paternal effects from effects caused by the integration of male genetic material; up to now, however, paternal effects remain largely unexplored in these systems. Here, we test for paternal effects in the gynogenetic Amazon molly ( Poecilia formosa ): a naturally clonal, all-female species with no parental care. Using a highly controlled breeding experiment involving 59 Atlantic molly males ( P. mexicana ) and 57 Amazon molly females, we generated 169 broods and 2,966 offspring. While males were drawn from a naturally variable stock population, females – next to being genetically identical – were highly standardized for age, size, descent, and developmental experience. We asked whether male identity or body size predicted offspring size – a key offspring phenotypic trait. We also asked whether male identity or body size predicted brood size. While we found substantial variation in both offspring size and brood size, we found no evidence for paternal effects on either trait. Next to providing an experimental test for paternal effects in a gynogenetic system, our results also strengthen the Amazon molly’s status as a model species for studying – in a highly controlled fashion – the developmental emergence of phenotypic variation. ### Competing Interest Statement The authors have declared no competing interest.

In many animal species, collective behaviours can be explained by a simple set of interaction rules. It is an intriguing question whether this generality at the level of mechanism also translates into generality at the level of function. Assuming that collective behaviour provides antipredator benefits for the partaking individuals, we ask whether the same collective behaviour provides protection against different predators in general. We investigated this question in sulphur-adapted fishes in their natural habitats in Mexico. Here, fish schools are frequently attacked by many different bird species and fish respond with synchronized and often repeated collective diving behaviour (escape waves). We found all bird species to wait longer until they attacked as they encountered more waves, both before they launched their first attack (pre-attack) and between subsequent attacks (post-attack). Post-attack, all bird species triggered similarly high numbers of waves while species differed in the number and the interval between the waves they triggered pre-attack. Through simulated bird arrivals, we confirmed that birds in the pre-attack context could be perceived as less threatening or completely overlooked, depending on their size, colouration and contrast to the background. We argue that the generality in the fish’s collective response as well as the similarity in effect on the different birds’ hunting behaviour might be explained by waves targeting a weakness in the visual processing ability common to different predators.

In collective biological systems, social contagion processes play an important role in evaluating and processing information on the level of the collective. These group-level abilities typically arise from individual-level mechanisms and through local interactions. We are interested in the role of these mechanisms and their effect on the system's response to environmental inputs. In this paper, we present a spatially embedded network model that is inspired by large fish shoals performing collective action in response to predation. We compare the observations of spatio-temporal dynamics in the model simulations with empirical observations, studying specifically the effect of spatial heterogeneities on system activity. The model demonstrates how already simple mechanisms suffice to represent key characteristics of the study system, and highlights the importance of taking into account the spatial embedding for understanding group-level processes in animal collectives.

Predation risk is one of the most important factors generating behavioral differences among populations. In addition, recent attention focusses on predation as a potential driver of patterns of individual behavioral variation within prey populations. Previous studies provide mixed results, reporting either increased or decreased among-individual variation in response to risk. Here, we take an explicit developmental approach to documenting how among-individual variation develops over time in response to predator exposure, controlling for both genetic and experiential differences among individuals. We reared juveniles of naturally clonal Amazon mollies, Poecilia formosa, either with or without a predator visible during feedings over 4 weeks and analyzed activity during feedings, time spent feeding and number of visits to the feeding spot. (I) Predator-exposed fish did not differ from control fish in average feeding behavior, but they were less active during feeding trials. (II) In the absence of the predator, substantial changes in among-individual variation over time were detected: among-individual differences in feeding duration increased whereas differences in activity decreased, but there were no changes in feeder visits. In contrast, in the presence of a predator, among-individual variation in all three behaviors was stable over time and often lower compared to control conditions. Our work suggests that predation risk may have an overall stabilizing effect on the development of individual variation and that differences in predation risk may well lead to population-wide differences in among-individual behavioral variation.

The Model AI Assignments session seeks to gather and dis- seminate the best assignment designs of the Artificial In- telligence (AI) Education community. Recognizing that as- signments form the core of student learning experience, we here present abstracts of five AI assignments from the 2024 session that are easily adoptable, playfully engaging, and flexible for a variety of instructor needs. Assignment spec- ifications and supporting resources may be found at http://modelai.gettysburg.edu.